International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
  
forest cover with different densities: 0, 25, 50, 70 and 100 %. 
The used ozone concentrations were related to the following 
zones: tropical (0.247 g.cm-2), subarctic summer (0.346 
g.cm-2), average latitudes in winter (0.395 g.cm-2) and 
subarctic in winter (0.48 g.cm-2). Furthermore, we used two 
extreme values of ozone contents corresponding to 0.495 
g.cm3 and 0.959 g.cm3. These concentrations were 
programmed in the 6S code database. For each simulation, we 
considered the three sensors and various forests cover 
densities. 
2.3 Spectral indices 
Theoretically, the “ideal” vegetation index should be sensitive 
to vegetation cover, insensitive to soil background (color, 
brightness, moisture and roughness), independent of the 
spatial and spectral resolutions of the sensors, little affected 
by atmospheric and environment effects, does not saturate 
rapidly, normalize the drift of the sensor radiometric 
calibration, as well as solar illumination geometry and senor 
viewing conditions (Jackson ef al, 1983; Bannari, 1996). 
These effects intervene simultaneously during in situ 
measurements and at the time of the satellite and/or airborne 
images data acquisition. In the literature, more than fifty 
vegetation indices (Bannari et al, 1995) were developed for 
different applications and to correct some of these different 
problems. In this study we retain only those that are 
developed to minimize the soil and atmospheric effects. 
The Normalized Difference Vegetation Index (NDVI) 
developed by Rouse et al. (1974) is the most popular index 
and the most used in various remote sensing applications. 
This index has undergone several transformations to minimize 
soil and atmospheric effects. By considering the bare soil line 
parameters, slope and origin, Richardson and Wiegand (1977) 
developed the Perpendicular Vegetation Index (PVI). 
However, Huete (1988) demonstrated that there was a 
contradiction between the NDVI and PVI indices in 
describing the spectral behaviour of vegetation and soil 
background. Consequently, he developed a new vegetation 
index called the Soil Adjusted Vegetation Index (SAVI), 
which is somewhat a compromise between ratio indices 
(NDVI) and orthogonal indices (PVI). The originality of this 
transformation lies in the establishment of a simple model, 
which describes adequately the soil-vegetation system. In 
order to reduce the soil color and brightness on the SAVI, 
Baret et al. (1989) proposed a new version of this index: the 
Transformed Soil Adjusted Vegetation Index (TSAVI). The 
soil line parameters (slope and origin) are introduced into the 
calculation of this index, which gives it a global character, i.e. 
it requires the use of only one index for different applications 
instead of using a determined index for each specific 
application (Baret ef al., 1989). To improve the sensitivity of 
SAVI to vegetation and to increase its potential to 
discriminate the bare soil, Qi et al (1994) proposed a 
modified version: the Modified Soil Adjusted Vegetation Index 
(MSAVI). Rondeaux er al. (1996) adapted the TSAVI 
especially for agricultural applications in a new version 
named Optimized Soil Adjusted Vegetation Index (OSAVI). 
The OSAVI is a particular case of the TSAVI when the slope 
(a) and the origin (b) of soil line are equal to 1 and 0, 
respectively 
In order to correct atmospheric diffusion on the NDVI, 
Kaufman and Tanré (1992) developed a new vegetation 
802 
index: the Atmospherically Resistant Vegetation Index 
(ARVI). A self-correction process for the atmospheric effect 
on the red channel accomplishes the resistance of this index to 
atmospheric effects. The resistance degree of the ARVI to the 
atmospheric variations depends on the accuracy of the 
determination of the atmospheric self-correction coefficient. 
Based on the 5S code, Kaufman and Tanré (1992) recommend 
the unit value for self-correction coefficient (y ^ 1) allow a 
better adjustment for most remote sensing applications; unless 
the aeroso| model is known a priori. To correct the 
atmospheric effects on the TSAVI, Bannari ef al. (1997) have 
proposed the Transformed Soil Atmospherically Resistant 
Vegetation Index (TSARVI). This transformation was based 
on the substitution of the red channel by the red-blue channel 
as suggested by Kaufman and Tanré (1992) and on the 
calculation of the bare soil line parameters (slope and origin) 
in the red-blue/NIR apparent spectral space. Developed 
especially for AVHRR sensor by using only apparent 
reflectances, the Global Environment Monitoring Index 
(GEMI) is a non-linear index. The objective of the GEMI is 
to evaluate and manage globally the environment without 
being affected by the atmosphere (Pinty and Verstraete, 
1992). For a combined correction of the atmospheric effects 
and optical properties of soil background, Huete ef a/. (1996) 
proposed a new version of SAVI named the Enhanced 
Vegetation Index (EVI). 
Theoretically, the values of the optimal vegetation index must 
be between 0 and |, respectively, for bare soil and dense 
vegetation cover. However, because of the disturbances and 
the problems raised above, the perfect linearity is not obtained 
by any vegetation index (Bannari ef a/., 2000). This problem 
is partially caused by the high sensitivity to the chlorophyll 
absorption in the red, which saturates very quickly (Huete et 
al., 1999). In order to solve the linearity problem, Roujean 
and Breon (1995) proposed the Renormalized Difference 
Vegetation Index (RDVI). This index is a simple re- 
normalization of the NDVI in order to have a very good linear 
relationship to the surface biophysics parameters. As for the 
Modified Simple Ratio (MSR), it is an improved version of the 
RDVI for biophysical parameters extraction in boreal forest 
environment (Chen, 1996). To solve the linearity problem and 
to correct atmospheric effects, Gitelson ef al. (1996) proposed 
the Green Atmospherically Resistant Vegetation Index 
(GARI), which exploits apparent reflectance in the blue, red, 
green, and near infrared channels. 
  
  
NDVI=Lrır — Pr) (1) 
(Prır +P r) 
py] 5s ap ib (2) 
Ja? «1 
SAP Pee Pe rr) (3) 
PrirtPrtL 
OSA yo oen an (4) 
(pnt pri 0,16) 
[2p melo pal] 8p PI DR ) 
MSAVI= 5 (5) 
Int. 
TS 
AI 
TS 
Wit 
GE 
Wit 
EV 
RI 
M. 
GL 
Whe 
pn: 
PPIR 
PB: 
Pp: 
pv: 
pr : 
P pir 
P RB 
Par: 
A b. 
y à 
aan 
red/! 
4, al 
the r 
Lis